Marvin

Especially `hot_allocation` in Zipfian workloads

While prefix bloom filters are pretty simple in theory, they need a Trait the user needs to implement. Possible API: ```rs trait PrefixExtractor { fn extract(&self, key: &[u8]) -> impl...

lz4_flex has support for dictionaries

https://github.com/facebook/rocksdb/wiki/Indexing-SST-Files-for-Better-Lookup-Performance

And possibly the amount of single delete + value pairs that *could* be removed if the GC watermark goes higher. That way, based on that number, we can trigger a...

Get rid of hard coded L1->L2 rewrite and instead lazily rewrite long-living segments. Forcing a rewrite into L2 was a monkey patch to make sure monotonic series do not end...

Currently the iterator syntax can not really represent key-value separation. For instance, if we wanted to get the list of keys of a prefix, using something like `tree.prefix(p).map(...)` would end...

Currently, when doing a range read over an uncached bunch of blobs, each blob will incur an `fopen()` syscall. If the blobs sit in the same blob file, the repeated...

and compare to bloom filters and blocked bloom filters